[OpenAFS-port-darwin] AFS crash

[Derrick J Brashear]

On Thu, 21 Oct 2004, Ragnar Sundblad wrote:

>
> --On den 8 oktober 2004 14:44 -0400 "Michael G. Ross" <mgross@csail.mit.edu> 
> wrote:
>
>> panic(cpu 1): Increase -stat parameter of afsd(VLRU cycle?)
>
> We got that panic too pretty easily, until we upped the
> -stat to 20000 in /private/var/db/openafs/etc/config/afsd.options.
>
> This has been good and well until today, when we started to distribute
> another 50 000 files to our machines with themis (like AFS' "package"
> but a little smarter, it mirrors trees of files onto the local disk).
>
> Setting -lstat to 30000 makes the machine panic soon after boot.
>
> Is there anything we can do to avoid this crash?
>
> Our afsd.options currently looks like so:
> -stat 20000 -dcache 2400 -daemons 5 -volumes 128 -dynroot -fakestat-all

I haven't seen this since fixing the vnode ref leak from the parent back 
in June. If you can give me a simple reproducible test case, I can fix it 
(but probably not for a few weeks). I still have a pending kernel crash 
dump from Steve Lidie to look at which may or may not tell me something 
about a different problem.